I am using some product descriptions to generate a content-based recommender system where I am using TF_IDF for generating vectors for descriptions and then calculating the similarity using cosine_similarity. Based on it, it's recommending products. Now, I have seen the LIME doc and a few tutorials on how LIME can be configured for text classifiers in this section. But I can't understand how I can configure LIME for my mentioned use case to see which words are affecting most in my recommender system. Is it even possible to do it using LIME?
For your convenience, here is a sample notebook of my recommender system.
I am using some product descriptions to generate a content-based recommender system where I am using TF_IDF for generating vectors for descriptions and then calculating the similarity using cosine_similarity. Based on it, it's recommending products. Now, I have seen the LIME doc and a few tutorials on how LIME can be configured for text classifiers in this section. But I can't understand how I can configure LIME for my mentioned use case to see which words are affecting most in my recommender system. Is it even possible to do it using LIME? For your convenience, here is a sample notebook of my recommender system.